EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
2174. Optimizing Treatment Allocation in the Presence of Interference: a Causal Machine Learning Approach to the Influence Maximization Problem
Invited abstract in session MC-31: Causal Machine Learning, stream Analytics.
Monday, 12:30-14:00Room: 046 (building: 208)
Authors (first author is the speaker)
1. | Daan Caljon
|
Faculty of Economics and Business, KU Leuven | |
2. | Jente Van Belle
|
LIRIS, KU Leuven | |
3. | Wouter Verbeke
|
Faculty of Economics and Business, KU Leuven |
Abstract
In Influence Maximization (IM), the objective is to select the optimal set of entities in a network to target with a treatment so as to maximize the total effect. For instance, in marketing, the objective is to target the optimal set of customers that maximizes the total response rate, resulting from both direct treatment effects on targeted customers and indirect, spillover effects that follow from targeting these customers.
In selecting the optimal set of entities, current IM methods typically overlook the features of the entities. Nevertheless, these features might contain important information on the propagation of effects through the network. Moreover, current methods assume that all entities that receive treatment will be activated (e.g., buy the product), ignoring varying susceptibilities to treatment depending on an entity’s characteristics.
Current IM approaches mainly rely on assumptions about how effects propagate through the network, potentially leading to a suboptimal selection of entities. To improve upon current methods, we propose a causal machine learning approach which estimates both direct and spillover effects from data, instead of relying on assumptions. After training a causal estimator on observational data, in a second step, treatment allocation is optimized by integrating the treatment effect estimates into current IM algorithms. This two-step approach is shown to improve the expected influence spread compared to traditional methods.
Keywords
- Analytics and Data Science
- Machine Learning
- Graphs and Networks
Status: accepted
Back to the list of papers